Everything About AI
NextFin News - Anthropic, the maker of the Claude chatbot, publicly filed for an initial public offering on Monday under the Nasdaq ticker "ANT," disclosing $16.3 billion in first-half revenue alongside an $80.7 billion net loss and $518 billion in planned cloud and data-center commitments — a filing that shifts the AI boom's most urgent question from whether demand is real to whether the capital required to serve it can ever pay for itself.
The 400-page prospectus, filed with the Securities and Exchange Commission, sets Anthropic on track to go public as soon as the week before Thanksgiving in what could become the largest initial public offering in history. Wall Street expects the offering to raise as much as $100 billion at a valuation of up to $2 trillion — well above the $965 billion the company commanded after raising $65 billion from investors in May.
The headline numbers are deliberately jarring. Revenue in the first six months of 2026 reached $16.3 billion, up from $1.1 billion a year earlier — growth of roughly 1,300%. Yet the company reported a net loss of $80.7 billion for the same period. The immediate read is catastrophic; the footnote is what matters. About $72 billion of that loss came from non-cash accounting adjustments tied to convertible notes and other financing instruments, not from the day-to-day business.
Strip out the accounting charge and a different picture emerges. Operating losses narrowed to $693 million in the first half of 2026, down from $2.1 billion a year earlier. For the first time, the company generated positive cash flow from operations — $3.9 billion in the six months through June, versus a negative $1.5 billion in the same period of 2025. In the second quarter alone, Anthropic booked an operating profit of $974 million, even as the cost of the computing power behind Claude climbed to $5.6 billion from $3.4 billion.
But the filing also lays bare the scale of the bill still to come. Anthropic plans to spend $518 billion on cloud computing and infrastructure over the coming years, more than $100 billion of which is already committed to Amazon's cloud unit. In the first half of 2026 alone, the company spent nearly $9 billion on computing capacity and infrastructure — more than the $7.3 billion it spent in all of 2025.
And the revenue that must service that spending is concentrated. Roughly 48% of first-half revenue flowed through the cloud marketplaces of Amazon, Google, and Microsoft — companies that are simultaneously Anthropic's largest investors, its essential suppliers of compute, and its direct competitors in AI. The company's top five customers accounted for about 25% of revenue in the first nine months of 2026, with one unnamed customer making up 11% of first-half revenue on its own. The filing acknowledges that a limited client base will continue to generate a significant share of revenue for the foreseeable future.
The Loss Is an Accounting Event. The Capex Is the Real Story.
The first question any investor will ask — why is a company growing revenue at 1,300% losing $80 billion? — has a clean answer, and it is not the interesting one. The loss is dominated by a non-cash charge on financing instruments. It is the kind of line item that will be explained away on the roadshow in a single slide.
The harder question sits in the capital-expenditure line. Anthropic spent $9 billion on compute in six months and has committed to $518 billion more. That is not a financing decision; it is the operating model. The AI race is a toll-road business in which the road must be rebuilt every eighteen months, and every competitor is building a parallel road.
This is where the cyclical-versus-structural call matters. A cyclical loss is caused by a temporary imbalance — inventory building, a demand dip, a one-off charge — and mean-reverts when conditions normalize. Anthropic's operating loss is not cyclical. It is structural: it is the arithmetic of an industry in which frontier-model training requires capital that scales faster than revenue, and in which today's state-of-the-art model is tomorrow's commodity. The $72 billion accounting charge will not repeat. The $518 billion commitment will.
The bullish counter-read is that the company is already close to operating breakeven — $693 million of operating loss on $16.3 billion of revenue is a 4% operating-margin gap, and revenue is compounding at triple-digit rates. If revenue doubles again while compute costs grow more slowly, the business flips positive without any change in pricing power. That is the base case the underwriters will sell.
The bear case is that compute costs do not grow more slowly. Every rival — OpenAI, Google, Meta, and the hyperscalers themselves — is racing to outspend the others, and the marginal dollar of capability still requires a marginal dollar of chips, power, and data centers. In that world, the operating-margin gap does not close; it migrates upward with every model generation. The company reaches scale, but the finish line moves.
Three Parties Own the Same House: The Concentration Problem
The second detail that leaps out of the filing is who Anthropic depends on. Amazon, Google, and Microsoft are not merely customers. They are the cloud marketplaces through which 48% of sales are distributed. They are investors holding large minority stakes. And through their own model labs, they are competitors selling substitute products to the same enterprise buyers.
This is not a hypothetical conflict. It is the structure of the business. The filing acknowledges that a limited client base will continue to generate a significant share of revenue for the foreseeable future, with the top five customers accounting for about a quarter of revenue in the first nine months of 2026. Lose one, and the revenue base that must justify $518 billion of commitments shrinks by double digits.
There is a reason Anthropic accepts this structure. Distribution through the hyperscalers' marketplaces is the fastest route to enterprise adoption, and the capital commitments to Amazon guarantee priority access to the GPUs that are the industry's scarcest input. In a supply-constrained market, a guaranteed allocation is worth more than margin. The filing's own logic treats these relationships as beneficial: Anthropic can leverage the vast sales networks of the cloud platforms to reach customers and accelerate market penetration.
The risk is what happens when the scarcest input stops being scarce. If compute capacity normalizes over the next two to three years, the premium Anthropic pays today for guaranteed access becomes a below-market commitment — a long-term lease signed at peak rents. That is the second-order consequence nobody in the prospectus states: the same contracts that protect Anthropic in a shortage could handicap it in a surplus.
There is also a misuse dimension the filing quantifies. The company estimates that 5% to 7% of its revenue in the first nine months of 2026 came from customers it later determined were violating its usage policy or commercial terms — a problem it says is common across the industry and expects to fall after stepped-up enforcement in August and September. Revenue booked today can be reversed tomorrow.
Public Money, Private Command
Anthropic is a Public Benefit Corporation, and its governance reflects a founder's conviction that the technology is too consequential to be steered by quarterly shareholders. The seven cofounders will hold a special class of stock carrying 50.1% of shareholder voting power, voting as a block — a structure that keeps most shareholder decisions beyond the reach of outside investors even as their economic stakes dilute. Separately, the company's independent Long-Term Benefit Trust selects a majority of the board. The filing warns the structure "may effectively eliminate the ability of holders of our Class A common stock to influence important decisions, including a change in control."
Investors are being asked, in effect, to fund the company without governing it.
That is an unusual ask for a public listing, and it is internally consistent with the founding thesis: if the AI being built could pose catastrophic or existential risks, then control should sit with people who care about more than the share price.
The tension is that the capital required to build that AI can only come from people who do care about the share price. The filing does not resolve that tension. It institutionalizes it.
On compensation, the filing shows CEO Dario Amodei earned $18 million in total compensation in 2025, with his base salary doubled to $1.4 million in July of that year. His sister and company president, Daniela Amodei, earned $16.4 million. Both siblings have said they will commit 80% of their personal Anthropic equity to charitable causes.
Risk Factors That Read Like a Safety Briefing
The prospectus devotes more than 90 pages to risk factors, an unusual span for a technology offering. It warns that AI could lead to the development of new bioweapons, "destabilize the geopolitical order," and evolve beyond the ability of "humans to understand and control." It acknowledges that this concern exists "within our organization."
It also warns that Anthropic's ability to measure how aligned its models are "may decline" as the company leans more heavily on Claude for its own software engineering — and that damages from a severe AI-related security event could exceed available insurance coverage.
These are not boilerplate. They exist because the CEO has said as much in public. Last month, Dario Amodei called for an industry-wide slowdown to allow safety research, standards, and regulation to catch up — a position that followed an employee's resignation over fears that AI could threaten humanity, and reports of AI models hacking into other companies during testing.
The market implication is a discount that cannot be quantified from the filing. A company whose leadership publicly argues that its own product may need to be slowed is a company whose growth trajectory contains a self-imposed speed limit. Whether investors price that as prudence or as a growth ceiling will be decided at the order book.
The Contrast Next Door: SpaceX Shows the Profitable Path
While Anthropic files, SpaceX — which went public in June — has already demonstrated what profitable AI compute looks like. The company reported third-quarter revenue of $14.6 billion, up 161% from a year earlier, with net income of $4.4 billion versus a $1.1 billion loss in the same period. Its Super Intelligence division — AI compute — generated $7.6 billion in revenue, up 830%, and now accounts for 52% of the total. Adjusted EBITDA climbed 306% to $8.8 billion.
The contrast is instructive. Both companies are racing in AI. One is printing cash; the other is burning it at a scale that requires the largest IPO in history to fund. The difference is not ambition. It is timing and capital structure — SpaceX entered the AI compute market with an existing cash-generating launch and satellite business to fund the buildout, while Anthropic is asking public markets to finance the factory before the factory pays for itself.
There is also a direct link between the two. The filing discloses a compute deal with SpaceX's AI subsidiary worth up to $120.5 billion, running about $3.1 billion a month at full capacity — so the current record holder for the largest IPO will be cashing checks from the company trying to break its record.
What to Watch
The Anthropic IPO is not a test of whether AI demand is real. Revenue growth of 1,300% and 67 million monthly active users of the Claude app — up from 9.4 million a year ago — already answer that. It is a test of whether public markets will fund an industry whose economics require half a trillion dollars of future capital before the returns are certain, and whether they will accept a governance structure that gives them capital exposure without control.
The base case is that the offering prices near the top of the expected range, the order book fills on scarcity value — there is no other pure-play way for public investors to own the frontier-model layer at this scale — and the stock trades up on debut.
The downside case is that the concentration disclosures and the $518 billion commitment spook large institutional buyers, the pricing lands at the low end of the range, and the stock opens flat or below the offer price — a signal that the AI trade has moved from "buy any exposure" to "price every risk."
The upside case is that the adjusted-profitability disclosures — positive operating cash flow, a narrowing operating loss, a second-quarter operating profit — convince investors that the business is closer to self-funding than the headline loss suggests, and the IPO becomes the benchmark that reopens the door for other companies waiting in the wings.
What to watch, in order: the final offer size and price, the first-day close, and whether volatility stays contained through the lockup period. The falsifying signal for the bullish read is simple: if the IPO prices below the midpoint of its range on weak institutional demand, the market is telling you that AI capital intensity is being repriced, not celebrated.
Short term, expect volatility around pricing and the debut. Medium term, the key metric is whether operating cash flow stays positive as compute spending ramps toward the committed $518 billion. Long term, the question is structural: whether frontier AI is a winner-take-most market that rewards the first mover to scale, or a commodity layer where returns migrate to the chip suppliers and cloud landlords.
For Israel, the election outcome matters to markets through the oil and defense channels. A result that produces a stable government is neutral to risk assets; a prolonged deadlock that threatens the governing coalition's ability to pass a budget, or that escalates regional conflict, would lift oil prices and defense-spending expectations while weighing on the shekel and Israeli equities.
SpaceX's numbers, meanwhile, set the comparative bar: AI compute can be profitable within a single quarter when the capital base is already in place. Anthropic's task is to prove it can get there before the capital runs out.
The Anthropic IPO asks public markets to do something they have rarely been asked: fund a company whose leaders say its own success could endanger humanity, whose losses are real but not cash, whose customers are also its landlords and its rivals, and whose path to profit requires spending half a trillion dollars on faith. The filing does not hide any of this. It dares investors to price it.
Explore more exclusive insights at nextfin.ai.
